Automatic velocity picking with convolutional neural networks
Yue Ma, Xu Ji, Tong W. Fei, Yi Luo · 2018
We developed an automatic velocity picking methodology based on convolutional neural networks (ConvNets). The proposed method formalizes the picking problem into a ConvNet regression model to map the NMO-corrected seismic gather to the velocity error estimates. We also propose a data preprocessing technique to normalize the shallow and deep reflections of a CMP gather into the same moveout shape, which is a key ingredient for successful training. A synthetic example shows the feasibility and effectiveness of the proposed method. Presentation Date: Wednesday, October 17, 2018 Start Time: 1:50:00 PM Location: 204B (Anaheim Convention Center) Presentation Type: Oral